Multiscale MD-AI framework for formulation property prediction and screening

Overview

The Multiscale MD-AI framework is an innovative solution designed to predict physicochemical properties of surfactant and polymer-based formulations. By integrating molecular dynamics (MD) simulations at various resolutions—atomistic, coarse-grained, and mesoscopic—with machine learning (AI), this framework offers a powerful tool for rapid virtual ingredient screening. It employs flexible parametrisation strategies, matching specific simulation methods to properties based on required spatiotemporal resolution. This approach not only enhances predictive accuracy but also accelerates the design and optimization of new formulations.

Technical specifications

Key features:

• Integrates multiscale MD simulations with AI for accurate property prediction
• Utilizes bottom-up and top-down parametrisation strategies for model optimization
• Capable of high-throughput screening for rapid virtual ingredient evaluation
• Validates predictive accuracy against experimental measurements of micellisation, film formation, and rheology

The framework leverages AI to explore design spaces and map properties efficiently, thereby supporting the development of new formulations that meet specific performance criteria. The combination of physical and data-driven methods allows for comprehensive screening across molecular to macroscopic scales.

Technology readiness level

Currently at Technology Readiness Level 3, the framework is in the proof-of-concept phase. Over the next 6-12 months, efforts will focus on developing standardized workflows for parameterizing smoothed dissipative particle dynamics (SDPD) simulations, fine-tuning coarse-grained parametrization, and implementing machine learning approaches for high-throughput screening. This phase will demonstrate the framework's predictive capabilities and prepare it for further development.